Case study · Shopify performance
The new store was live. On phones, customers watched it fall apart.
Days after KOR-E moved to Shopify, images and videos stopped loading for shoppers on mobile data, and only on mobile data. Six days later the store was dependable again.
kor-e.co.il
61 → 95
Mobile performance score
−60%
Lighter homepage
6 days
From first message to a store that worked
10 → 0
Accessibility problems
01 · The brief
A brand-new store, failing in front of customers.
KOR-E had just finished moving from WordPress to Shopify. Then the messages started arriving: on mobile data, customers could not see the furniture they had come to buy. Images failed. Videos stayed black. Whole sections looked like they had simply given up. On Wi-Fi, everything was fine.
That inconsistency is worse than a plainly broken site. The team could not reproduce it at their desks, could not tell which shopper would hit it next, and could not trust their new store to do its job the moment someone stepped outside.
The brief was a single sentence: make the media load properly on mobile. The fee was tied to that result, if it did not work on a real phone, there was nothing to pay.
02 · The diagnosis
The first explanation was convincing. It was also wrong.
The early evidence pointed at the mobile networks themselves. There was a real, measurable reason to believe certain Israeli carriers were throttling the kind of connection the store’s CDN preferred, and it explained perfectly why the failure only ever appeared on cellular.
A screen recording from the client showed what the measurements had missed. The store was asking for all 21 of its videos at the same moment. On a phone connection that wall of requests swallowed the entire pipe. The videos never finished, and because almost none of them had a still frame to fall back on, they simply stayed black. Everything below them was left starving. Wi-Fi just had enough room to hide it.
That changed everything. The store was not at the mercy of a mobile carrier. The cause was inside the theme, where it could actually be fixed.
Before, the live store on a throttled phone connection
After, the rebuilt store, same connection
Both panels are full-page captures taken under the same throttled connection, scroll either one to compare. Shown exactly as captured.
03 · What changed
Make it dependable first. Then take out the weight.
Sixty-two separate changes were measured. Five were undone when the measurements said they were not helping. What stayed made the store lighter, calmer and far more predictable on a phone.
01
Customers stopped competing with 21 videos at once
Videos now wait their turn until they are nearly on screen. On a slow connection the homepage video went from never playing at all to starting in 14 seconds.
02
Phones stopped downloading a hover effect they can’t use
Every product card was quietly loading a second image for a mouse-hover effect that only exists on desktop. On a phone that was half of all product-image data, spent on something no shopper could ever trigger.
03
The image library became an asset instead of a liability
126 files audited and 116 rebuilt. Twenty-one of the 28 “icons” turned out to be photographs in disguise, one of them, the size of a fingernail on screen, weighed 941 KB on its own.
04
Inherited code stopped slowing the first impression
The stylesheet was cut by more than a third against what the pages actually use, and a block of code sitting in the page was moved somewhere Shopify could compress it, 24.8 KB down to 3.8 KB.
05
The page stopped moving under the shopper’s thumb
Text no longer jumps as fonts load, and ten accessibility problems were fixed along the way.
Before
After
Time for the main image to appear, category page
7.4 seconds
5.1 seconds
Homepage media, fully scrolled
17.8 MB
7.2 MB
Accessibility problems
10
0
The hover-image finding was the one that paid for itself twice over: removing images no phone can ever display was pure gain, and desktop was left byte-for-byte identical.
04 · Where it landed
From 61 to 95 on mobile performance.
The store’s mobile performance score went from 61 before the work to 95 after it. More to the point, customers on ordinary mobile connections could reach the content again, and the client had a tested store instead of a launch problem that appeared at random.
Every contracted target is listed with its measured result, exactly as it was reported to the client. Where a number came in short, it is closed with a measured reason rather than an excuse.
Target
Goal
Result
Performance, five runs in a row
90+
95 / 92 / 93 / 91 / 91
met
Responsiveness to taps
under 150 ms
about 21 ms
met
Speed of first impression
under 2.5 s
about 2.05 s
met
Accessibility
98+
100
met
SEO
100
100
met
Layout stability
perfect zero
0.0006 · the fonts make zero unreachable
measured limit
Main image appears
under 2.5 s
2.586 s, 86 ms short
measured limit
First text appears
under 1.5 s
about 2.0 s, the platform floor is 1.8 s
measured limit
Best practices
100
77 · two cookies Shopify sets itself
vendor-owned
How this was measured: Lighthouse on mobile against the live site, medians over 5–16 runs with the ranges recorded, plus same-session before/after comparisons for individual changes. No real-world visitor data existed for this store, so nothing here is presented as a field measurement. Real Mobile Safari on a real cellular connection was never tested, there was no physical handset, and that gap is stated rather than papered over.
05 · What the client said
Sent by WhatsApp after the engagement closed.
Is your store breaking in ways you can’t reliably reproduce?
I’ll follow your customer’s actual experience to the real cause, fix what is fixable, and give you the evidence, and the calm, to move on.
Book a call
More work
More stores, built the same way.
